Piper Morgan - AI Product Management Assistant

Started Out Vibe Coding, Soon Hit the Harder Stuff

What started as a curiosity about AI-powered product management has turned into something unexpected: a systematic methodology for human-AI collaboration that actually works.

I'm Christian Crumlish (you can also call me xian), and I've been documenting every breakthrough, every failure, and every 'streamlist command not found' moment of building Piper Morgan—an AI assistant that aims not to replace PM judgment but to amplify it through transparent, systematic collaboration.

This isn't another AI tool promising to automate your job away. It's a learning journey shared in real-time, with all the technical debt and environment setup comedy included.

What we're doing

Building in public

When I started this project in May 2025, I had a simple question: Could AI actually help product managers work more systematically, or was this all just hype cycle noise?

Within a week I was able to use my prototype to accomplish a simple but meaningful task in my workplace (refining a task description and writing a comment explaining the changes).

Turns out, the answer depends entirely on how you approach the collaboration. Skip the systematic thinking, and you get ChatGPT with extra steps. Embrace verification-first development and transparent documentation, and something genuinely useful starts to emerge.

Discovered so far…

Excellence Flywheel methodology

A systematic approach to AI-human collaboration that scales from 15-minute task automation to complex architectural decisions.

Multi-agent coordination

Strategic deployment of specialized AI tools (Claude Code, Cursor, Opus) assigned focused roles with clear separation of concerns and crisp handoff protocols to overcome the “seam” in available tooling today. No more prompt chaos or minimal context loss.

Verification before implementation

Every pattern gets tested, every assumption gets checked. Sounds slow, actually accelerates development once you stop debugging assumptions.

GitHub-first tracking and transparency

All work tracked with clear acceptance criteria. Zero architectural drift across 50+ implementations because the process forces explicit decision-making.

Current reality check

Where we actually are

This is a learning project that has grown wings. We're not a startup, we're not taking funding, and we're definitely not claiming to have solved product management.

What we have done is discover repeatable patterns for AI-augmented PM work that maintain human judgment while systematically capturing and sharing what works.

Recent breakthroughs include:

  • • Multi-agent orchestration baked into workflows with negligible overhead
  • • Integration of systematic verification into rapid development cycles
  • • Agent coordination patterns that prevent the usual AI collaboration chaos and feed recursively into Piper's education

What's next: Continuing to build in public, sharing what we learn, and discovering whether these patterns scale beyond one learning PM and his robot programming assistants.

For technical users: All development work is documented with full architectural decisions, implementation details, and systematic methodology at pmorgan.tech — comprehensive technical documentation for developers who want to understand how these patterns actually work.

Why follow this journey?

For founders, senior product people, and UX designers and strategists

If you're curious about AI augmentation that respects human expertise rather than trying to replace it, this project documents patterns you can adapt to your own context.

  • • Real implementation decisions with full context
  • • Failures and course corrections, not just success stories
  • • Systematic approaches to human-AI collaboration
  • • How to maintain quality when everything moves faster

For architects, lead developers, unicorns, and other makers

We're documenting not just what we build, but how we think about building it. Every session logged, every architectural decision explained, every comedy-of-errors shared. We also routinely survey our logs to detect new patterns emerging, and we continually experiment with context management and methodology reinforcement and enforcement (checking your work).

  • • Transparent process documentation
  • • Decision frameworks for AI tool selection
  • • Patterns for maintaining systematic excellence under pressure
  • • Evidence that environment setup struggles are universal

For other AI-curious professionals

Watch how systematic thinking translates AI potential into practical value, without the usual hype cycle noise.

Get involved

Follow the learning

Building Piper Morgan (LinkedIn Newsletter • 659 subscribers): Daily blog posts narrating the development and architectural design process blow by blow with updates on discoveries, breakthroughs, and the occasional technical debt confessional. (Blog posts lag behind development work by about six weeks now and growing.) Weekly Ships updating progress in real time. Plus, announcements. Read or (subscribe to) the newsletter

Building Piper Morgan (Medium Publication): For the impatient, blog posts appear on Medium roughly a week after development and a good six weeks before they make it to LinkedIn (currently), with the last 20 and all process or insight posts always free to read. Read the series

The Pygmalion Effect (Rosenverse): A free talk I gave in August 2025 (requires signup) for a primary audience of mid-level and senior UX practitioners, covering the seductive appeal of vibe coding, information architecture saves everything, the Excellence Flywheel, multi-agent coordination patterns, your role isn't disappearing—it's evolving, IA principles for AI system design, from "should designers code?" to systematic AI assistance, the human advocacy imperative, the Bootstrap Moment ahead, what this means for your career, and systematic kindness. See the talk

This Site: Methodology documentation, pattern catalog, and the occasional deep dive into what we're learning. For those for whom the blog is a mite too technical (too much Python code!), a new serial we’re working on coming soon: Growing Piper Morgan, with more broadly applicable insights about effectively working with LLMs today on the orchestration frontier.

Questions?

This is an experiment in transparent development and systematic learning. If you're curious about specific patterns, want to share similar experiences, or just want to follow along, the best way to stay connected is through the newsletter.

We're learning in public because the best discoveries happen when smart people share what actually works.

Want to dig deeper? Check out How It Works or follow along with the blog..